PRACTICAL

STATISTICS
FOR
MEDICAL
RESEARCH
Douglas G. Altman
Head ofMedical Statistical Laboratory
Imperial Cancer Research Fund
London

CHAPMAN & HALL/CRC

Boca Raton London New York Washington, D.C.
Contents

Preface xi

1 Statistics in medical research 1

1.1 Statistics at large 1
1.2 Statistics in medicine 3
1.3 Statistics in medical research 4
1.4 What does statistics cover? 5
1.5 The scope of this book 8

2 Types of data 10
2.1 Introduction 10
2.2 Categorical data 10
2.3 Numerical data 11
2.4 Other types of data 13
2.5 Censored data 16
2.6 Variability 17
2.7 Importance of the type of data 17
2.8 Dealing with numbers 17

3 Describing data 19
3.1 Introduction 19
3.2 Averages 21
3.3 Describing variability 22
3.4 Quantifying variability 31
3.5 Two variables 38
3.6 The effect of transforming the data 41
3.7 Data presentation 42
Exercises 45

4 Theoretical distributions 48
4.1 Introduction 48
4.2 Probability 49
4.3 Samples and populations 50
4.4 Probability distributions 50
4.5 The Normal distribution 51
 Contents vii
4.6 The Lognormal distributions 60
4.7 The Binomial distribution 63
4.8 The Poisson distribution 66
4.9 Mathematical calculations 68
4.10 The Uniform distribution 71
4.11 Concluding remarks 71
Exercises 71

5 Designing research 74
5.1 Introduction 74
5.2 Categories of research design 75
5.3 Sources of variation 78
5.4 An experiment: is the blood pressure the same in both arms? 79
5.5 The design of experiments 80
5.6 The structure of an experiment 83
5.7 Random allocation 85
5.8 Minimization 91
5.9 Observational studies 91
5.10 The case-control study 93
5.11 The cohort study 96
5.12 The cross-sectional study 99
5.13 Studies of change over time 101
5.14 Choosing a study design 102
Exercises 103

6 Using a computer 107

6.1 Introduction 107
6.2 Advantages of using a computer 107
6.3 Disadvantages of using a computer 108
6.4 Types of statistical program 110
6.5 Evaluating a statistical package 111
6.6 Strategy for computer-aided analysis 112
6.7 Forms for data collection 114
6.8 Plotting 119
6.9 Other uses of computers 120
6.10 Misuses of the computer 120
6.11 Concluding remarks 121

7 Preparing to analyse data 122

7.1 Introduction 122
7.2 Data checking 122
7.3 Outliers 126
7-4 Missing data 130
7.5 Data screening 132
vni Contents
7.6 Why transform data? 143
7.7 Other features of the data 146
7.8 Concluding remarks 149
Exercises 149

8 Principles of statistical analysis 152

8.1 Introduction 152
8.2 Sampling distributions 153
8.3 A demonstration of the distribution of sample means 155
8.4 Estimation 160
8.5 Hypothesis testing 165
8.6 Non-parametric methods 171
8.7 Statistical modelling 173
8.8 Estimation or hypothesis testing? 174
8.9 Strategy for analysing data 175
8.10 Presentation of results 176
8.11 Summary 177
Exercises 177

9 Comparing groups - continuous data 179

9.1 Introduction 179
9.2 Choosing an appropriate method of analysis 179
9.3 The t distribution 181
9.4 One group of observations 183
9.5 Two groups of paired observations 189
9.6 Two independent groups of observations 191
9.7 Analysis of skewed data 199
9.8 Three or more independent groups of observations 205
9.9 One way analysis of variance - mathematics and
worked example 218
9.10 Presentation of results 220
9.11 Summary 222
Exercises 223

10 Comparing groups - categorical data 229

10.1 Introduction 229
10.2 One proportion 230
10.3 Proportions in two independent groups 232
10.4 Two paired proportions 235
10.5 Comparing several proportions 241
10.6 The analysis of frequency tables 241
10.7 2 x 2 frequency tables - comparison of two proportions 250
10.8 2 x k tables - comparison of several proportions 259
10.9 Large tables with ordered categories 265
 Contents ix
10.10 k x k tables - analysis of matched variables 266
10.11 Comparing risks 266
10.12 Presentation of results 271
10.13 Summary 271
Exercises 272

11 Relation between two continuous variables 277

11.1 Association, prediction and agreement 277
11.2 Correlation 278
11.3 Use and misuse of correlation 282
11.4 Rank correlation 285
11.5 Adjusting a correlation for another variable 288
11.6 Use of the correlation coefficient in assessing non-Normality 291
11.7 Correlation - mathematics and worked examples 293
11.8 Interpretation of correlation 297
11.9 Presentation of correlation 300
11.10 Regression 300
11.11 Use of regression 306
11.12 Extensions 309
11.13 Regression - mathematics and worked example 311
11.14 Interpretation of regression 316
11.15 Relation to other analyses 318
11.16 Presentation of regression 319
11.17 Regression or correlation? 320
Exercises 321

12 Relation between several variables 325

12.1 Introduction 325
12.2 Analysis of variance and multiple regression 325
12.3 Two way analysis of variance 326
12.4 Multiple regression 336
12.5 Logistic regression 351
12.6 Discriminant analysis 358
12.7 Other methods 360
Exercises 361

13 Analysis of survival times 365

13.1 Introduction 365
13.2 Survival probabilities 367
13.3 Comparing survival curves in two groups 371
13.4 Mathematical calculations and worked examples 377
13.5 Incorrect analyses 385
13.6 Modelling survival - the Cox regression model 387
13.7 Desien of survival studies 393
x Contents
13.8 Presentation of results 393
Exercises 394

14 Some common problems in medical research 396

14.1 Introduction 396
14.2 Method comparison studies 396
14.3 Inter-rater agreement 403
14.4 Diagnostic tests 409
14.5 Reference intervals 419
14.6 Serial measurements 426
14.7 Cyclic variation 433
Exercises 435

15 Clinical trials 440

15.1 Introduction 440
15.2 Design of clinical trials 441
15.3 Sample size 455
15.4 Analysis 461
15.5 Interpretation of results 471
15.6 Writing up and assessing clinical trials 473
Exercises 474

16 The medical literature 477

16.1 Introduction 477
16.2 The growth of statistics in medical research 478
16.3 Statistics in published papers 481
16.4 Reading a scientific paper 493
16.5 Writing a scientific paper 498
Exercises 499

Appendix A Mathematical notation 505

Al.l Introduction 505
A1.2 Basic ideas 505
A1.3 Mathematical symbols 509
A1.4 Functions 510
A1.5 Glossary of notation 510

Appendix B Statistical tables 514

Answers to exercises 546

References 575
Index 589